Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Iometer measures storage performance under a workload you define; it does not produce one universal “disk speed” score. You choose the target, read/write mix, access pattern, transfer size, concurrency and run length, then interpret IOPS, throughput, latency and errors in that context. Before testing, decide whether to use a dedicated empty physical disk or a bounded test file on a volume. Raw-device tests can destroy data, so never select a disk containing files you need.
Iometer’s project site describes the tool as an I/O workload generator and measurement tool. Its established concepts remain useful, though its available user guide uses older Windows terminology. Verify the package and executable compatibility for your system rather than assuming every build works on every current Windows version.
What Iometer measures
Iometer generates storage I/O according to an access specification and reports how the tested subsystem behaves under that pattern. It can model sequential or random access, reads, writes or a mixture, different transfer sizes, multiple workers and targets, and local or distributed workloads. It can also save configurations and write results to CSV. That flexibility makes it useful for controlled comparisons and approximating known workloads, but the results only describe the conditions you tested.
A 4-KiB random-read result, for example, does not predict large sequential write performance. Nor does a synthetic peak necessarily represent an application. Record the workload settings alongside every result.
#1 Best Overall
- Easily store and access 2TB to content on the go with the Seagate Portable Drive, a USB external hard drive
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop
- To get set up, connect the portable hard drive to a computer for automatic recognition no software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
Before you run a test
- Obtain an Iometer package containing
Iometer.exeandDynamo.exe, and check its architecture and compatibility with your Windows version. The project history and current project information are at iometer.org; do not infer a current release number from the older guide. - Use a dedicated, empty physical disk for raw-device characterization, or a bounded test file on a volume when existing data must remain intact.
- Back up important data and verify the selected target twice. Avoid the OS disk unless it is intentionally disposable and the procedure specifically requires it.
- Confirm required permissions. Raw physical-device access may require administrator rights. Check that the target is writable and not being used by applications or background tasks.
- Record whether storage is local, USB-attached, network-attached, virtualized, RAID/SAN-backed or cloud-backed. These layers affect what the test measures.
- Note antivirus, indexing, encryption, snapshots, deduplication, compression, thin provisioning and other activity that may affect results.
Iometer terminology
- Iometer: The graphical controller and test coordinator.
- Dynamo: The workload generator. Starting Iometer normally starts a local Dynamo instance automatically.
- Manager: A Dynamo instance representing a machine.
- Worker: A thread within a manager that performs I/O.
- Target: The physical disk, logical volume or test file receiving I/O.
- Access specification: The pattern of requests, including transfer size, read/write mix and random/sequential mix.
- Outstanding I/Os: The maximum asynchronous requests a worker attempts to keep active per selected disk. Actual device queue depth may be lower.
For a remote manager, the legacy guide gives dynamo IOServer as an example, where IOServer is the Iometer machine name. Exact command-line syntax and executable names may differ by package. The guide says one Dynamo process per machine is sufficient; add workers within that instance as needed.
Choose raw-device or file-based testing
Physical disk
The guide indicates physical drives as PHYSICALDRIVE:n when the drive contains nothing but free space. Raw-device testing is useful for characterizing a dedicated device with fewer filesystem effects, but it is destructive: it can overwrite partitions, filesystems and data. Confirm the device identifier, disconnect or stop other users of it, and never select a disk with data you need. A RAID volume, SAN LUN, virtual disk or cloud volume may not expose an underlying physical disk at all.
Logical volume
For a logical target, Iometer uses a file named iobw.tst. The guide says the program creates and grows this file during preparation or when testing begins if it does not already exist. This is generally safer than raw-device testing when a volume contains other data, but only if the intended volume has adequate free space, the file is the intended target and the test is bounded. It still measures the combined effects of the filesystem, metadata, allocation and alignment, caching, encryption, virtualization and any storage services in the path. File-based and raw-device figures are not interchangeable.
Configure a basic test
- Extract the package and keep
Iometer.exeandDynamo.exetogether. A remote workload machine also needs access toDynamo.exe. - Launch
Iometer.exe. In the Topology panel, select the local manager. - Open Disk Targets, identify the intended disk or logical volume, and verify again that it is the correct target. If a manager’s device list appears stale, the guide says to right-click the manager to update its target lists.
- For a logical volume, ensure the test file can be prepared and decide its size and location before running. Do not use an unbounded or production target casually.
- Open Access Specifications. Edit or duplicate a specification and set the transfer size, read/write distribution and random/sequential distribution to match the question you are asking.
- Set the target area, starting sector and outstanding I/O count. The guide’s Maximum Disk Size is in 512-byte sectors; zero means the full disk or test file beginning at the starting sector.
- Choose a finite run time. Where appropriate, allow a warm-up period or exclude initial unstable measurements from analysis.
- Open Results Display, choose the statistics to show and set an update interval. Start the test and select a results filename if prompted.
- After the run, save the test configuration as an
.icffile and retain the CSV output with your system and workload notes.
Labels and exact dialog behavior can vary by build. The Iometer guide’s quick-start sequence provides additional detail on selecting a manager, targets, access specifications and result display.
Rank #2
- Easily store and access 5TB of content on the go with the Seagate portable drive, a USB external hard Drive
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop
- To get set up, connect the portable hard drive to a computer for automatic recognition software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
Design a workload that answers a question
Do not treat an access specification as a speed setting. It defines the workload. Important controls include request size, read/write and random/sequential percentages, outstanding I/Os, worker count, target size, starting sector, run time and any target open/close behavior.
| Question | Illustrative workload | What to keep in mind |
|---|---|---|
| What sequential bandwidth is available? | 128 KiB or 1 MiB requests; test 100% reads and 100% writes separately; 100% sequential; compare queue depth 1 with a higher point. | Use the transfer size required by your application or comparison. The legacy guide’s 64-KB sequential-read example is historical, not a universal standard. |
| How does random I/O scale? | 4-KiB requests; test 100% reads, 100% writes and a defined mixed case separately; 100% random; for example, queue depths 1, 4, 16 and 32. | This is an example test matrix, not a required standard. Small-block results are highly workload- and device-dependent. |
| How will a particular application perform? | Use measured application transfer-size distribution, read/write ratio, access pattern, concurrency and working-set size. | Do not infer a workload from a label such as “database.” Validate the synthetic profile against application-level behavior on an isolated test system or non-production copy. |
The legacy guide’s default access specification is documented as 2-KB random I/O, 67% reads and 33% writes, described as database-like. Treat this as a historical example, not a modern database standard. Its 512-byte sequential-read example is likewise not a guarantee of maximum IOPS.
Queue depth is total concurrency, not just one setting
Iometer’s outstanding-I/O value is a per-worker, per-disk maximum. The total potential workload multiplies across workers and disks. For example:
4 workers × 2 disks per worker × 16 outstanding I/Os = up to 128 outstanding I/OsThe documented default is 1. Increase gradually and monitor the system: very high totals can overwhelm a Windows storage driver or available memory, causing stalls, hangs or crashes. For cloud volumes, select queue depth based on the workload and provisioned performance, not an arbitrary high number; AWS EBS guidance emphasizes workload-dependent tuning.
Rank #3
- Easily store and access 1TB to content on the go with the Seagate Portable Drive, a USB external hard drive.Specific uses: Personal
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop. Reformatting may be required for Mac
- To get set up, connect the portable hard drive to a computer for automatic recognition no software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
Pick an appropriate test area
A test area smaller than available cache may mostly measure cache behavior. A larger area can better expose sustained media performance, but its appropriate size depends on the device and purpose. SSDs may deliver a short burst before thermal throttling or dynamic write-cache exhaustion; HDD performance can vary by platter location. Thin-provisioned storage may allocate blocks on writes, and reusing an old test file can differ from testing a fresh one. Starting-sector changes can also change alignment. Keep test area, starting sector and preparation policy consistent when comparing runs.
Run, save and repeat
For a meaningful comparison, keep the machine, controller, power profile, firmware, driver, filesystem state, target area, request size, read/write mix, access pattern, worker count, queue depth and duration constant. Record temperature and thermal conditions, background activity and the idle or preconditioning period. A useful sequence is:
- Confirm the target and stop competing workloads where practical.
- Record system and storage conditions.
- Run an appropriate preparation or warm-up phase. State whether the reported result includes it.
- Run at least three measured repetitions.
- Exclude a failed or anomalous run only for a documented reason.
- Report the average and spread, not only the best result, and retain the configuration and CSV.
Distinguish burst performance from sustained behavior, and both from application performance. Short runs can emphasize cache; longer ones may reveal thermal limits, garbage collection or other sustained effects. A deliberately favorable synthetic pattern is useful for a peak measurement but is not an application forecast.
Automate with a saved configuration
The legacy guide documents batch-mode examples such as:
Rank #4
- Easily store and access 4TB of content on the go with the Seagate Portable Drive, a USB external hard drive.Specific uses: Personal
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop
- To get set up, connect the portable hard drive to a computer for automatic recognition no software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
iometer /c bigtest.icf /r bigtest_results.csvIt also lists iometer /r out.csv, iometer /c test.icf /r results.csv and iometer /c test.icf /r results.csv /t 100. In the documented behavior, /c loads a configuration and /r names the results file. Batch mode runs the configured tests, writes results and exits. The /t option controls how long Iometer waits for managers; it is not the workload duration. Set a nonzero run time in the saved test configuration, since a zero duration can leave a run continuing indefinitely. Verify these options against the particular package in use.
Interpret the results
- IOPS: completed I/O operations per second. Look at read and write rates separately when the workload mixes them.
- Throughput: data transferred per unit of time. As a rough relationship, throughput ≈ IOPS × transfer size, but units and mixed workloads affect reported values.
- Latency or response time: time taken to complete I/O. Average latency alone can hide slow outliers, so examine time-series behavior and, where available, latency distributions.
- Errors: any I/O failures make a performance number suspect and should be investigated.
- Aggregation: distinguish worker, manager and total results. Confirm that targets were assigned as intended and that workers are not unintentionally duplicating or sharing them.
Do not report only MB/s: a device can have strong sequential bandwidth but weak small-block random latency, or the reverse. State whether units are MB/s or MiB/s when the reporting convention is known. Compare results between Iometer, fio, DiskSpd or another tool only when effective workload, target, buffering, duration, concurrency and units are equivalent.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Troubleshoot common problems
The disk does not appear
Possible causes include a partitioned or data-bearing disk not being exposed as a raw physical target, an offline disk, an unwritable logical volume, a stale manager target list, controller/driver behavior or a virtual machine exposing only a virtual disk. The guide says physical drives are shown only when entirely free space and logical drives only when writable. Refresh the manager’s target list and confirm the device through the operating system before proceeding.
A logical target has a red slash
This generally indicates that iobw.tst needs preparation. Check free space, write permission, file locks, target selection and whether the proposed bounded test belongs on a system or production volume.
Best Value
- [Upgraded Version] - This external hard drive features a mirrored logo stripe combined with a striped anti-slip design, and the rounded corners of the casing make it easier to grip. The stripes also have a heat dissipation function, ensuring stable and fast data transfer.
- 【Ultra-thin and quiet】 - The motherboard adopts JMicron 578 noise-free solution, giving you a quiet working environment. Lightweight and portable size designed to fit in your pocket for easy portability.
- 【Ultra-Fast Data Transfers】 - Pairing this external hard drive with JMicron 578 solution USB 3.0 and USB 2.0 interfaces enables blazing-fast data transfer. It boasts theoretical read speeds of up to 125MB/s and write speeds of up to 103MB/s.
- 【Plug and Play】 - With no software to install, just plug it in and the drive is ready to use.The hard disk chip is wrapped with an aluminum anti-interference layer to increase heat dissipation and protect data.
- 【What You Get】 - 1 x Portable Hard Drive, 1 x USB 3.0 Cable, 1 x User Manual, Gift-type shell packaging ,Three-year manufacturer's warranty and free technical support services.
The run hangs or crashes
Reduce outstanding I/Os first, then worker count, number of targets and test-file size. If needed, reconsider unusually large transfer sizes. The combined concurrency can exceed what a driver or system can handle.
Results are unexpectedly high
Check for a test area smaller than cache, buffered I/O, sparse or thin-provisioned files, controller write-back cache, short run time, pre-throttling burst behavior, unintended worker/target assignments, or cloud burst-credit effects.
Results vary between runs
Check background tasks, antivirus and indexing, temperature, power management, SSD garbage collection, RAID rebuilds, cloud throttling, test-file freshness, manager/worker assignments and warm-up policy. Retain the spread across repeated runs rather than selecting the most flattering number.
The CSV is missing or incomplete
The guide says Iometer may prompt for a results filename when a run starts if none was supplied. It also documents that specifying a results file on the command line records results even if the GUI setting is None. Confirm the output path and that the run completed.
When to use another tool
Iometer is a reasonable choice when you need a GUI workload builder, saved access specifications, multiple workers and targets, distributed coordination or compatibility with an existing Iometer procedure. Its older documentation and package-dependent behavior make it less attractive for new, heavily automated workflows.
- fio: A cross-platform, job-file-driven tool with broad workload controls, scripting, client/server operation, logging and percentile reporting. See the fio documentation. The guide’s illustrative job below must be adapted to the system and target;
libaioand direct I/O support are not universal.
[global]
ioengine=libaio
direct=1
runtime=60
time_based=1
ramp_time=10
size=10G
filename=/path/to/testfile
group_reporting=1
[randread]
rw=randread
bs=4k
iodepth=16
numjobs=1- Microsoft DiskSpd: A Windows-focused command-line load generator useful for scripted tests. Microsoft’s repository and releases list version 2.2 dated June 3, 2024; its release notes warn that asynchronous I/O loop changes require re-baselining results at queue depths above 1.
- CrystalDiskMark: A simpler consumer-oriented benchmark for quick checks, not detailed application workload modeling. AWS lists it alongside fio and DiskSpd in its EBS benchmarking guidance.
Reproducibility record
Include these details with every published or shared result:
Quick Recap
- Target device or test-file path; raw or logical; capacity and tested area
- Read/write ratio, random/sequential ratio and transfer size
- Starting sector, worker count, targets per worker and outstanding I/Os
- Run duration, warm-up/preconditioning policy and number of repetitions
- IOPS, bandwidth, latency, errors and whether figures are per worker or aggregate
- Hardware, OS, storage controller, driver, firmware, filesystem and relevant cache conditions
- Temperature, power profile, background workload and whether the device was burst or sustained
- Iometer package/build, saved
.icffile and CSV results
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
